THE Hacker News reports on the growing risk posed by AI agents operating within organisations, potentially with broad access to internal systems and credentials, yet with little to no oversight. The article frames such agents as insiders in effect: if an employee had that level of reach with minimal supervision, it would be treated as an insider risk.
It highlights a forthcoming webinar, AI as an Insider Threat: From Shadow AI to Governed AI, featuring Richard De La Torre of Bitdefender, promised to deliver a practical security playbook for bringing unmanaged AI under control.
Key findings from Bitdefender’s 2026 Cybersecurity Assessment and related observations are cited to illustrate visibility gaps. The piece notes that only 51.8% of organisations report full visibility into the AI tools their staff use, leaving nearly half without a complete picture. IBM is referenced as finding that 20% of breached organisations were compromised through shadow AI.
The article summarises the webinar’s practical aims: identifying where AI appears on endpoints, presenting real-world research from Bitdefender Labs that includes a malicious agent observed in the wild, and discussing why blocking AI outright often fails while governed adoption offers a more workable path. It emphasises that an unmanaged AI agent with credentials and no owner behaves like an insider risk and outlines the suggested governance-based approach to reduce that risk.